3D multimodal dental dataset based on CBCT and oral scan

该数据集包含了290名健康受试者的扫描图像，其中287份被选中以确保数据维度的一致性。在为VQ-GAN模型训练做准备的过程中，这些图像经过了降采样和填充的预处理步骤。具体来说，该数据集被划分为217份用于训练，70份用于推断。预处理还包括将体素大小降采样至2毫米立方体，并对重建任务进行随机裁剪。该数据集的任务是医疗成像中的异常检测与分割。

This dataset comprises scan images from 290 healthy subjects, with 287 samples selected to ensure consistent data dimensions. In preparation for training the VQ-GAN model, these images underwent preprocessing procedures including downsampling and padding. Specifically, the dataset is partitioned into 217 training samples and 70 inference samples. Additional preprocessing steps involve downsampling the voxel size to 2 mm cubic voxels and applying random cropping for the reconstruction task. The core task of this dataset is anomaly detection and segmentation in medical imaging.

搜集汇总

背景与挑战

背景概述

该数据集提供了一个预训练的nnU-Net模型，专门用于牙颌面CBCT和CT扫描的分割，支持5个特定类别的分割任务，包括上颅骨、下颌骨、上牙、下牙和下颌管。数据集还包含了模型的使用指南和相关研究论文的引用信息。